Spatio-Temporal Modelling of Road Traffic Fatalities in the Western Cape

Abstract

A machine learning model capable of predicting the probability of road fatal events in both time and space is developed. By aggregating relevant features of the Western Cape into an H3 grid, the model learns patterns of fatal events. Traditional machine learning and deep learning techniques are employed to understand the relationship between aggregated features and road fatal events. The models are compared against each other and against historical average models currently used in the industry. This study represents the first attempt to use machine learning techniques to model road traffic fatalities in South Africa, specifically in the Western Cape. 

https://doi.org/10.59200/ICONIC.2024.016